CoChat

Out-group animosity drives engagement on social media

Open access
Rathje, Steve, Van Bavel, Jay J., van der Linden, Sander · 2021 · Proceedings of the National Academy of Sciences

The DOI matches its Crossref record; 75 of the 75 references it cites were found in a registry or an index; 13 of the 61 sources its claims cite were checked for support.

Abstract

= 2,730,215), we found that posts about the political out-group were shared or retweeted about twice as often as posts about the in-group. Each individual term referring to the political out-group increased the odds of a social media post being shared by 67%. Out-group language consistently emerged as the strongest predictor of shares and retweets: the average effect size of out-group language was about 4.8 times as strong as that of negative affect language and about 6.7 times as strong as that of moral-emotional language-both established predictors of social media engagement. Language about the out-group was a very strong predictor of "angry" reactions (the most popular reactions across all datasets), and language about the in-group was a strong predictor of "love" reactions, reflecting in-group favoritism and out-group derogation. This out-group effect was not moderated by political orientation or social media platform, but stronger effects were found among political leaders than among news media accounts. In sum, out-group language is the strongest predictor of social media engagement across all relevant predictors measured, suggesting that social media may be creating perverse incentives for content expressing out-group animosity.

Overview

CoChat Summary

TL;DR

Across 2,730,215 Facebook posts and tweets from news media and US congressional accounts, political out-group language was the strongest predictor of sharing and retweeting. The out-group effect was much larger than in-group language or emotional language, and out-group posts especially elicited “angry” reactions.

Key findings

  • In an internal meta-analysis across eight datasets, each political out-group word increased the odds of a retweet or share by about 67% (exp(b) = 1.67, 95% CI = [1.43, 1.69], P < 0.001), whereas political in-group language had no statistically significant effect (exp(b) = 1.05, 95% CI = [0.90, 1.22], P = 0.563).
  • The out-group effect was larger than established emotional predictors: the percent increase in shares associated with out-group language was about 4.8 times that of negative affect language and about 6.7 times that of moral-emotional language. Meta-analytically, negative affect increased diffusion by about 14% per word, moral-emotional language by 10%, and positive affect decreased diffusion by about 5%.
  • Posts with at least one out-group word were more than twice as likely to be shared as posts with at least one in-group word (exp(b) = 2.32, 95% CI = [1.57, 2.47], P < 0.001; a 132% increase).
  • Out-group language strongly predicted “angry” reactions (meta-analytic exp(b) = 2.19, 95% CI = [1.68, 2.84], P < 0.001), while in-group language strongly predicted “love” reactions (exp(b) = 1.57, 95% CI = [1.24, 2.00], P < 0.001). The “angry” reaction was the most popular reaction in seven of eight datasets.

Methods

  • Observational big-data study of public Facebook posts and tweets from liberal and conservative news media accounts and Democratic and Republican members of Congress, totaling n = 2,730,215 observations. Study 1 media data included 599,999 Facebook posts and 227,229 tweets; Study 2 congressional data included 1,078,562 tweets and 825,424 Facebook posts.
  • Text was analyzed with dictionaries for liberal/conservative in-group and out-group terms, positive and negative affect, and moral-emotional language. Ordinary least squares regressions predicted log-transformed retweet/share counts, controlling for URL, media, follower count, and retweet status, followed by random-effects internal meta-analyses across the eight datasets.

Limitations

  • The paper states the results are correlational, not causal, and calls for experimental work to determine causation and clarify why out-group content is engaging.
  • The paper states data access limitations prevented following up important questions such as who is producing the engagement.
  • The paper states the data likely reflect a specific time frame—the years leading up to the 2020 election—so generalizability to other time periods or nations is unclear, and the role of platform algorithms is unclear because companies are not transparent about them.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1073/pnas.2024292118

Cites

75 of 75 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Fig. 1. Political out-group words were the strongest predictors of shares and retweets from both liberal and conservative news media accounts (A and B, respectively) and liberal and conservative congress member accounts (C and D, respectively) on Facebook and Twitter. By comparison, political in-group words, as well as measures of discrete emotions, such as positive emotion words, negative emotion words, and moral-emotional words, were relatively weak predictors of shares and retweets. Error bars represent 95% CI (though error bars are small).
Fig. 1. Political out-group words were the strongest predictors of shares and retweets from both liberal and conservative news media accounts (A and B, respectively) and liberal and conservative congress member accounts (C and D, respectively) on Facebook and Twitter. By comparison, political in-group words, as well as measures of discrete emotions, such as positive emotion words, negative emotion words, and moral-emotional words, were relatively weak predictors of shares and retweets. Error bars represent 95% CI (though error bars are small).CC BY · PubMed Central
Fig. 2. In-group and out-group words predicted different types of engagement in both the liberal and conservative news media accounts (A and B, respectively) and congress accounts (C and D, respectively). Political out-group words were strong predictors of shares, comments, “haha,” and “angry” reactions, whereas in-group words were strong predictors of “love” reactions. Reactions are shown as they are shown on the Facebook and Twitter platforms (from Top to Bottom: share, comment, like, heart, haha, wow, sad, angry, retweet, and favorite). Error bars represent 95% CI (though error bars are small).
Fig. 2. In-group and out-group words predicted different types of engagement in both the liberal and conservative news media accounts (A and B, respectively) and congress accounts (C and D, respectively). Political out-group words were strong predictors of shares, comments, “haha,” and “angry” reactions, whereas in-group words were strong predictors of “love” reactions. Reactions are shown as they are shown on the Facebook and Twitter platforms (from Top to Bottom: share, comment, like, heart, haha, wow, sad, angry, retweet, and favorite). Error bars represent 95% CI (though error bars are small).CC BY · PubMed Central
Fig. 3. In A, the meta-analyzed effect sizes across all eight datasets are shown. The effect size of political out-group language was 4.7 times as large as that of negative affect language and 6.7 times as large as moral-emotional language. In B, the meta-analyzed effect sizes of in-group and out-group language on each of the Facebook reactions are shown. In-group language predicted “love” reactions, whereas out-group language strongly predicted “angry” reactions.
Fig. 3. In A, the meta-analyzed effect sizes across all eight datasets are shown. The effect size of political out-group language was 4.7 times as large as that of negative affect language and 6.7 times as large as moral-emotional language. In B, the meta-analyzed effect sizes of in-group and out-group language on each of the Facebook reactions are shown. In-group language predicted “love” reactions, whereas out-group language strongly predicted “angry” reactions.CC BY · PubMed Central

Verification

Claims the paper makes

12 Valid · 0 Not valid · 31 Not checked · 104 Own results and interpretations

73 more claims

Create a free account to see all 147 claims and to verify the 3 references still open.

See all claims — free account
  • Probably not a substantial claim. 40
  • It may state two separate claims as one. 36

Create a free account to see these claims.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

8 more claims about this paper

Create a free account to see every claim other papers make about this one.

See all claims — free account

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

2 of 2

Citations

Cites

Show all 75 references

Research works printed but not linked: 2

  1. A. Campbell, P. E. Converse, W. E. Miller, D. E. Stokes, The American Voter New York (Wiley, 1960).
  2. A. Mosseri, News Feed FYI: Bringing People Closer Together (Facebook Newsroom, 2018).

Cited by

See all 584

Metadata

Authors
Rathje, S., Van Bavel, J. J., & van der Linden, S.
Affiliations
  • University of Cambridge (GB)
  • Department of Psychology, Center for Neural Science, New York University, New York, NY 10003
  • Department of Psychology, University of Cambridge, Cambridge CB2 3RQ, United Kingdom;
Corresponding author
Rathje, Steve
Venue
Proceedings of the National Academy of Sciences
Published
2021-06-23
Volume
118
Issue
26
ISSN
  • 0027-8424 (ISSN-L)
  • 1091-6490
OpenAlex
W3176626547
Cited by
560
References
72
Citations by year
2026: 103 · 2025: 126 · 2024: 133 · 2023: 106 · 2022: 72 · 2021: 18
FWCI
108.57
Type
journal article
Language
en
Open access
Locations
  • Proceedings of the National Academy of Sciences (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • Apollo (University of Cambridge) (submittedVersion, cc-by)
  • OSF Preprints (OSF Preprints) (submittedVersion)
  • Apollo (cc-by)
Funders
  • Gates Cambridge Trust (OPP1144)
  • John Templeton Foundation (61378)
Topics
  • Social Sciences › Communication › Social Media and Politics
  • Social Sciences › Sociology and Political Science › Misinformation and Its Impacts
  • Social Sciences › Sociology and Political Science › Social and Intergroup Psychology
Keywords
social media engagement, in-group favoritism, outgroup derogation, political polarization, social media sharing, retweets, political language, political leaders, political orientation
MeSH terms
Emotions, Humans, Politics, Databases as Topic, Social Media

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Out-group animosity drives engagement on social media. https://app.cochat.ai/papers/10.1073/pnas.2024292118
The paper
Rathje, S., Van Bavel, J. J., & van der Linden, S. (2021). Out-group animosity drives engagement on social media. Proceedings of the National Academy of Sciences. https://doi.org/10.1073/pnas.2024292118

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.